Financial-Grade Cloud Native Architecture and Data Intelligence Practices from Ant Group

Ant Group (formerly Ant Financial) reflects on fifteen years of technology transformation, emphasizing five foundational technologies—Blockchain, Artificial Intelligence, Security, IoT, and Cloud Computing—as the basis for financial innovation.

The core challenge is computational capability, addressing both OLTP (online transaction processing) and OLAP (online analytical processing) to support massive payment volumes and intelligent data services.

Financial‑grade cloud native solutions rely on distributed micro‑services, message queues, and especially robust distributed transactions, high availability, and strong consistency. Ant’s open‑source middleware SOFAStack, with over 23,000 stars, exemplifies these practices.

Ant’s proprietary distributed database OceanBase, built on Paxos for strong consistency, now powers all internal data workloads, achieving peaks of 42 million transactions per second, clusters of over 1,000 nodes, storage beyond 2 PB, and tables with more than 320 billion rows while maintaining RPO = 0 and RTO < 30 seconds.

To ensure zero data loss and continuous operation, Ant employs a multi‑region, multi‑center active‑active architecture, demonstrated by a live server‑cut test that recovered services within 25 seconds.

Given the high frequency of software changes (up to 300,000 per year), Ant has built a technical risk‑as‑a‑service (TRaaS) platform that detects and recovers from failures within five minutes, complemented by proactive fault‑injection (red‑blue team) exercises.

Moving toward a mesh‑based infrastructure, Ant abstracts distributed middleware, databases, and disaster‑recovery capabilities into a Service Mesh, simplifying application development by offloading financial‑grade concerns to the underlying platform.

On the data intelligence side, Ant addresses PB‑scale data processing with internal MaxCompute, streaming engines like Flink, graph processing via GeaBase, and OLAP/MPP solutions such as ADS and Explorer, while recognizing challenges of heterogeneous frameworks, storage costs, and governance.

The proposed open compute architecture includes a unified storage layer, pluggable compute engines via SPI, a standardized SQL access layer, and an integrated data‑intelligence development platform that reduces code size dramatically.

For complex use cases like real‑time anti‑fraud detection, Ant has built a fusion compute framework that unifies streaming, graph, and machine‑learning workloads on a dynamic graph meta‑engine, later extended with the Ray engine for multi‑modal processing.

SQLFlow, an open‑source AI‑oriented SQL extension, abstracts end‑to‑end data‑to‑model pipelines, making AI development as simple as writing SQL.

Ant’s talent development program, BASIC College, focuses on the five BASIC domains (Blockchain, AI, Security, IoT, Cloud Computing) to cultivate expertise in foundational computer‑science skills essential for future architectural evolution.